Robotic task planning

Abstract

It is recognized herein that current approaches to autonomous operations are often limited to grasping and manipulation operations that can be performed in a single step. It is further recognized herein that there are various operations in robotics (e.g., assembly tasks) that require multiple steps or a sequence of motions to be performed. To determine or plan a sequence of motions for fulfilling a task, an autonomous system that includes a robot can perform object recognition, pose estimation, affordance analysis, decision-making, probabilistic task or motion planning, and object manipulation.

What is claimed is: 
     
         1 . A method for operating an autonomous machine in a physical environment, the method comprising:
 detecting an object within the physical environment;   performing pose estimation on the object so as to determine an initial state of the object;   identifying a task that requires that the autonomous machine interact with the object;   based on the task, determining a goal state of the object;   determining a plurality of intermediate states associated with the object, the intermediate states defining respective motion sequences for the object to reach the goal state from the initial state;   selecting one of the motion sequences, so as to define a selected motion sequence; and   the autonomous machine performing the selected motion sequence, thereby fulfilling the task.   
     
     
         2 . The method as recited in  claim 1 , the method further comprising:
 performing the pose estimation so as to determine a pose associated with the goal state of the object.   
     
     
         3 . The method as recited in  claim 2 , the method further comprising:
 determining the plurality of intermediate states based on the initial state, the pose associated with the goal state, and the task.   
     
     
         4 . The method as recited in  claim 3 , wherein determining the plurality of intermediate steps further comprises:
 performing an affordance analysis on the object so as to determine a plurality of feasible actions for the autonomous machine in completing the task.   
     
     
         5 . The method as recited in  claim 1 , wherein determining the plurality of intermediate states further comprises:
 generating an affordance map associated with the object,   based on the affordance map, determining that the goal state of the object is not reachable without reaching at least one of the plurality of intermediate states.   
     
     
         6 . The method as recited in  claim 5 , the method further comprising:
 based on the affordance map and task, generating at least one intermediate state of the plurality of intermediate states, the at least one additional intermediate state enabling the goal state to be reachable; and   augmenting the affordance map with the at least one additional intermediate state.   
     
     
         7 . The method as recited in a  claim 1 , wherein selecting one of the motion sequences comprises:
 determining that the selected motion sequence defines a path that is shorter than the other motion sequences.   
     
     
         8 . The method as recited in  claim 5 , wherein selecting one of the motion sequences further comprises:
 solving a Markov decision problem defined by the initial state, the goal state, and the plurality of the intermediate states.   
     
     
         9 . The method as recited in  claim 1 , wherein the selected motion sequence defines a plurality of state transitions, and performing the selected motion sequence further comprises:
 executing a first state transition of the plurality of state transitions; and   before executing a second state transition of the plurality of state transitions that directly follows the first state transition, determining whether a first intermediate state associated with the first state transition is reached.   
     
     
         10 . An autonomous system, the autonomous system comprising:
 an autonomous machine configured to operate in a physical environment;   a processor; and   a memory storing instructions that, when executed by the processor, cause the autonomous system to:
 detecting an object within the physical environment; 
 perform pose estimation on the object so as to determine an initial state of the object; 
 identify a task that requires that the autonomous machine interact with the object; 
 based on the task, determining a goal state of the object; 
 determining a plurality of intermediate states associated with the object, the intermediate states defining respective motion sequences for the object to reach the goal state from the initial state; and 
 select one of the motion sequences, so as to define a selected motion sequence, 
 wherein the autonomous machine is further configured to perform the selected motion sequence, so as to fulfill the task. 
   
     
     
         11 . The system as recited in  claim 10 , the memory further storing instructions that, when executed by the processor, further cause the autonomous system to:
 perform the pose estimation so as to determine a pose associated with the goal state of the object.   
     
     
         12 . The system as recited in  claim 11 , the memory further storing instructions that, when executed by the processor, further cause the autonomous system to:
 determine the plurality of intermediate states based on the initial state, the pose associated with the goal state, and the task.   
     
     
         13 . The system as recited in  claim 12 , the memory further storing instructions that, when executed by the processor, further cause the autonomous system to:
 perform an affordance analysis on the object so as to determine a plurality of feasible actions for the autonomous machine in completing the task.   
     
     
         14 . The system as recited in claim  14 , the memory further storing instructions that, when executed by the processor, further cause the autonomous system to:
 generate an affordance map associated with the object,   based on the affordance map, determine that the goal state of the object is not reachable without reaching at least one of the plurality of intermediate states.   
     
     
         15 . The system as recited in  claim 14 , the memory further storing instructions that, when executed by the processor, further cause the autonomous system to:
 based on the affordance map and task, generate at least one intermediate state of the plurality of intermediate states, the at least one additional intermediate state enabling the goal state to be reachable; and   augmenting the affordance map with the at least one additional intermediate state.   
     
     
         16 . The system as recited in  claim 10 , the memory further storing instructions that, when executed by the processor, further cause the autonomous system to:
 determine that the selected motion sequence defines a path that is shorter than the other motion sequences.   
     
     
         17 . The system as recited in  claim 14 , the memory further storing instructions that, when executed by the processor, further cause the autonomous system to:
 solve a Markov decision problem defined by the initial state, the goal state, and the plurality of the intermediate states.   
     
     
         18 . The system as recited in  claim 10 , wherein the selected motion sequence defines a plurality of state transitions, and the memory further stores instructions that, when executed by the processor, further cause the autonomous system to:
 execute a first state transition of the plurality of state transitions; and   before executing a second state transition of the plurality of state transitions that directly follows the first state transition, determine whether a first intermediate state associated with the first state transition is reached.